The Cybercab Hidden Joystick and the Robotaxi Race It Cannot Win on Tech Alone

A few days into the Cybercab’s public service, a passenger in Austin found something on the touchscreen that was not supposed to be there. Matthew Skrzypczak boarded his Cybercab, and instead of the usual passenger interface, the display showed a virtual joystick for manual driving. He tried it, filmed it, and posted it on X. The clip is thirty seconds long. But it touches a much bigger question than one strange ride: what does “fully autonomous” actually mean when the car still needs a human to move it?

The Joystick That Wasn’t Supposed to Exist #

On September 7, Skrzypczak posted a video of his Cybercab in what he called an “unusual state.” The single touchscreen was split in two. On the left sat a driving interface built around an on-screen joystick, with prompts to close the cabin doors and buttons for opening the door and calling support. On the right, live feeds from the car’s cameras — except the side cameras appeared frozen on an earlier frame from a Supercharger, suggesting the car had recently been moved by a Tesla employee.

“When I got in the vehicle, the display was in a different configuration with a virtual joystick for manual driving! I have a feeling I wasn’t supposed to see this… but I try to manually drive the vehicle anyway!” — Matthew Skrzypczak, on X, September 7

Why a No-Control Car Still Needs a Driver #

The explanation is almost mundane. A Cybercab has no steering wheel and no pedals, but sometimes the vehicle has to be repositioned by hand — nudged across a parking lot, or eased out of a tight spot when its route is blocked. Tesla never publicly demonstrated the joystick mode, yet it is exactly the kind of backdoor a robotaxi company needs before the software can handle every edge case on its own.

What Electrek Thinks #

The site that first covered the story has a sharp read on it. For creeping a car forward at walking speed, an on-screen joystick might be acceptable — but it looks clumsy, and it does not answer the question regulators actually care about: when an autonomous vehicle needs human intervention, is there a practical way for a human to take over? Tesla removed the wheel and pedals to make a statement that full autonomy has arrived. Yet the robotaxi service has been running on Model Ys for more than a year using the same autonomous hardware and software, and the Cybercab is, as Electrek puts it, essentially a different body on the same brain.

A Design That Left the Door Open #

What makes the joystick less surprising is that Tesla has been careful never to close the door on controls. Chairwoman Robyn Denholm said back in October that if regulators required it, the Cybercab could be fitted with conventional controls: “If we have to install a steering wheel, it can have a steering wheel and pedals.” Prototypes with wheels and pedals have been spotted in testing for years. The production car dropped them, but the platform clearly still knows how to accept them.

The Regulatory Question Nobody Solved #

The joystick news landed on the same day that the National Highway Traffic Safety Administration opened an inquiry into how Tesla certified a vehicle with no steering wheel, no pedals, and no mirrors — a review that covers roughly 1,000 cars. Rules were written around the assumption that a person sits behind a wheel, and Tesla’s argument that some of those standards do not apply to a fully driverless car is precisely what NHTSA wants to examine. Amazon’s Zoox has already navigated this path: in July it received approval for its steering-wheel-less robotaxi, becoming one of the first companies in the United States allowed to charge passengers in a vehicle with no wheel and no driver cabin. Beyond regulation, there is public perception. As an analyst at Gasgoo Research puts it, removing the wheel and pedals breaks consumer habits, and combined with the ongoing debate over Tesla’s camera-only approach, public acceptance will depend on real operating data and a long safety record — neither of which exists yet at meaningful scale.

The Race That Started Without Tesla #

The Cybercab finally carries passengers, but the market it entered is not the empty field Tesla imagined two years ago. Waymo operates fully driverless rides in 14 U.S. cities, has logged more than 220 million driverless miles as of March, and added Denver, San Diego, and Tampa earlier this month. Its latest safety data claims injury-causing crashes 82 percent below the human-driver baseline and serious-or-worse injuries 94 percent below. Zoox is running in Las Vegas and San Francisco, preparing to enter Austin and Miami, and recently expanded testing to Houston and San Diego. Registration numbers tell the same story: Tesla has 45 Cybercabs on the books in Texas, and about 420 robotaxi-class vehicles in total (mostly Model Y), while Waymo’s Texas fleet approaches 1,000. Tesla began its Austin robotaxi trial in June 2025 with Model Ys and promised rapid expansion; the rollout has been slower than the company’s own timeline.

Vision vs. Sensor Fusion #

Underneath the fleet numbers sits a philosophical divide. The Cybercab reads the road with eight cameras, feeding raw photon data through an end-to-end neural network — a bet that better data and stronger models can replace extra sensors. Waymo and Zoox combine cameras with lidar, radar, and high-definition maps, arguing that in a car with no human to catch a failure, redundant sensors that cross-check each other are the safer bet. Waymo has been blunt about it: in a summary of its 200 million miles of full autonomy, it said cameras alone are not enough. Neither approach is provably safer today; what Tesla lacks is the volume of driverless operating data needed to make the case.

Scale Is the Real Test #

Tesla’s counterargument has always been scale. The Cybercab runs FSD (Unsupervised) v15, removes the entire human-control system, and cuts the parts count sharply. Tesla has said the Austin factory can build up to 125,000 Cybercabs a year — a number that dwarfs the entire Waymo and Zoox fleets combined. Morgan Stanley’s Andrew Percoco framed the stakes simply: if Cybercab is mostly a launch with few cars on the road, expect the stock to feel pressure; if it scales and meaningfully expands the fleet, investor sentiment should improve. For a company that has promised full autonomy for a decade, “we are operating” is no longer enough of a proof. The joystick hidden in the menu is a small detail, but it is a reminder that the gap between the vision and the operating reality is still being crossed one edge case at a time. The first week of rider reports and the story of why Tesla built the Cybercab fill in the rest of the picture, alongside the official specifications.